NextMotionQA - Captioning: leaderboard
NextMotionQA (2026) tests how well vision-language models understand human motion shown as rendered 3D motion clips, with expert-verified items stratified by body part, direction, action and difficulty. This board is free-form captioning of 396 clips, rated 0 to 100 by a Qwen3.6-Plus judge for content coverage and action consistency; higher is better.
Metric: Task 2 captioning score (0-100): Qwen3.6-Plus judge rating of content coverage and action consistency of free-form captions of 396 rendered AMASS human-motion clips given to the model as RGB video, mean of the easy, medium and hard subsets; higher is better. Source: arxiv.org. Saturation forecast: Around 2031. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.5 27B | 35.57 | #359 |
| 2 | Qwen 3.6 Plus | 35.23 | #222 |
| 3 | Qwen 3.5 4B | 34.87 | #1175 |
| 4 | Qwen 3.5 9B | 33.73 | #803 |
| 5 | GPT-5.4 Mini | 32.77 | #389 |
| 6 | Qwen 3.5 0.8B | 23.3 | #8081 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=nextmotionqa-captioning · How It Works · Data refreshed daily, snapshot 2026-11-13.